OVATE FAMILY PROTEIN4 (OFP4) interaction with KNAT7 regulates secondary cell wall formation in Arabidopsis thaliana


(fax +1 684 822 6089; e-mail carl.douglas@ubc.ca)


The homeodomain transcription factor KNAT7 has been reported to be involved in the regulation of secondary cell wall biosynthesis. Previous work suggested that KNAT7 can interact with members of the Ovate Family Protein (OFP) transcription co-regulators. However, it remains unknown whether such an OFP–KNAT7 complex could be involved in the regulation of secondary cell wall biosynthesis in Arabidopsis. We re-tested OFP1 and OFP4 for their abilities to intact with KNAT7 using yeast two-hybrid assays, and verified KNAT7–OFP4 interaction but found only weak interaction between KNAT7 and OFP1. Further, the interaction of KNAT7 with OFP4 appears to be mediated by the KNAT7 homeodomain. We used bimolecular fluorescence complementation to confirm interactions and found that OFP1 and OFP4 both interact with KNAT7 in planta. Using a protoplast transient expression system we showed that KNAT7 as well as OFP1 and OFP4 act as transcriptional repressors. Furthermore, in planta interactions between KNAT7 and both OFP1 and OFP4 enhance KNAT7’s transcriptional repression activity. An ofp4 mutant exhibited similar irx and fiber cell wall phenotypes as knat7, and the phenotype of a double ofp4 knat7mutant was similar to those of the single mutants, consistent with the view that KNAT7 and OFP function in a common pathway or complex. Furthermore, the pleiotropic OFP1 and OFP4 overexpression phenotype was suppressed in a knat7 mutant background, suggesting that OFP1 and OFP4 functions depend at least partially on KNAT7 function. We propose that KNAT7 forms a functional complex with OFP proteins to regulate aspects of secondary cell wall formation.


Plant secondary cell walls are complex matrices consisting primarily of cellulose, hemicelluloses, and lignin, and are deposited during the maturation of specialized cell types including tracheary elements and fibers and are a considerable sink for fixed carbon derived from photosynthesis (Pauly and Keegstra, 2008). In the Arabidopsis thaliana (Arabidopsis) inflorescence stem, extensive secondary wall deposition occurs during the maturation of xylem vessels and xylary fiber cells and interfascicular fiber cells. This provides a useful model system to investigate both genes that contribute to secondary wall biosynthesis and genes that regulate the coordinated expression of multiple biosynthetic genes (Ehlting et al., 2005; Zhong and Ye, 2007). This system has been exploited to reveal a network of transcription factors regulating secondary cell wall biosynthesis (Zhong and Ye, 2007). Accumulating evidence shows that Arabidopsis NAC transcription factors such as SND1 (Zhong et al., 2006) and VND7 (Yamaguchi et al., 2008) act as master regulators of secondary cell wall biosynthesis that activate the developmental program of secondary cell wall biosynthesis both by directly targeting the regulation of cell wall biosynthetic genes and by regulating the expression of downstream transcriptional regulators of the MYB and homeodomain classes (Zhong et al., 2007, 2008; McCarthy et al., 2009; Zhou et al., 2009).

KNAT7 (At1g62990), one of seven Arabidopsis KNOTTED ARABIDOPSIS THALIANA (KNAT) genes and among the four Arabidopsis Class II KNAT paralogs (Kerstetter et al., 1994), has been shown to be one of the direct targets of both SND1 (Zhong et al., 2008) and MYB46 (Ko et al., 2009). KNAT7 was first identified as a regulator of secondary wall biosynthesis due to its co-expression with secondary cell wall biosynthetic enzymes (Brown et al., 2005; Ehlting et al., 2005; Persson et al., 2005) and due to the irregular xylem (irx) phenotype of a knat7 loss-of-function mutant (Brown et al., 2005) which led to its designation as IRX11 (Brown et al., 2005). In addition to the knat7 loss-of-function irx phenotype, expression of an engineered dominant transcriptional repression variant of KNAT7 in transgenic Arabidopsis results in thinner interfascicular cell walls (Zhong et al., 2008). However, other work shows that in addition to the irx phenotype, knat7 loss-of-function mutants have an increased cell wall thickness of interfascicular fibers, while KNAT7 overexpression mutants have the opposite phenotype (Li, 2009). One possible explanation for these apparently contradictory results is that KNAT7 acts as a transcriptional repressor rather than an activator in regulating secondary wall biosynthesis. In this case, the dominant transcriptional repression variant of KNAT7 would enhance rather than eliminate KNAT7 function. This possibility has not been explored in previous reports (Zhong et al., 2008).

KNOX and BELL-LIKE HOMEODOMAIN (BLH) proteins belong to the THREE AMINO ACID LOOP EXTENSION (TALE) family of plant homeodomain proteins that are well documented to interact in many plant species (Bellaoui et al., 2001; Muller et al., 2001; Smith et al., 2002; Chen et al., 2003). KNOX–BLH interactions form heterodimers that are thought to constitute functional complexes that regulate plant development (Bellaoui et al., 2001; Muller et al., 2001). The plant MEINOX domain consists of two smaller domains, KNOX1 and KNOX2, separated by a poorly conserved linker sequence, and the MEINOX domain is necessary and sufficient for interaction with BLH proteins (Bellaoui et al., 2001; Muller et al., 2001; Smith et al., 2002; Bhatt et al., 2004; Kumar et al., 2007). Systematic analysis of protein interactions of Arabidopsis TALE homeodomain proteins using the yeast two-hybrid system revealed a highly connected, complex network of interacting KNOX, BLH, and OVATE FAMILY PROTEINS (OFPs) in Arabidopsis, including interactions between KNAT7 and several OFPs including OFP1 and OFP4 (Hackbusch et al., 2005). These data provide the outlines of a possible TALE homedomain–OFP protein–protein interaction network. However, most of these interactions have not been validated by other tests of protein–protein interaction, and there is little information concerning the biological functions of OFPs alone or in proposed complexes with TALE homeodomain proteins.

OFP proteins belong to a plant-specific and poorly characterized family of plant regulatory proteins (Hackbusch et al., 2005). The tomato OVATE gene was originally found to encode a protein with a putative nuclear localization signal and an approximately 70-amino-acid C-terminal domain that is conserved in tomato, Arabidopsis, and rice (Liu et al., 2002). There are 18 genes in the Arabidopsis genome that encode proteins with a conserved OVATE domain, and most members of this family contain a predicted nuclear localization signal but lack recognizable DNA-binding domains (Hackbusch et al., 2005; Wang et al., 2007). Arabidopsis OFP1 has been shown to function as a transcriptional repressor and has a role in regulating cell elongation (Wang et al., 2007), while OFP5 was shown to negatively regulate the activity of a BLH1–KNAT3 complex during early embryo sac development in Arabidopsis (Pagnussat et al., 2007), providing evidence that KNAT–BLH–OVATE complexes may play regulatory roles in plants.

Here we report the functional characterization of Arabidopsis OFP4 (At1g06920), an OFP family member relatively closely related to OFP1 (At5g01840). We demonstrate that OFP1 and OFP4 interact with KNAT7 in vivo and that both OFP4 and KNAT7 are transcriptional repressors. We further show that the pleiotropic OFP1 and OFP4 overexpression phenotypes depend on KNAT7 function, and that an ofp4 mutant and ofp1 ofp4 double mutants have irx phenotypes similar to that of knat7. Taken together, these results indicate that OFP4 plays a role in regulating secondary cell wall formation through its interaction with KNAT7, expanding the very limited information regarding the functions of OFP proteins in plants.


KNAT7 interaction with OFP1 and OFP4

Investigation of a KNAT–BELL–OFP protein interaction network in a large-scale yeast two-hybrid screen showed the potential for KNAT7 to interact with a number of partner proteins including OFP1, OFP2, OFP3, OFP4, and OFP6 (Hackbusch et al., 2005). A review of microarray data over the course of inflorescence stem development (Ehlting et al., 2005) revealed that OFP1 is differentially regulated, with strongest expression in the oldest part of the stem (7–9 cm; up-regulated). Among the 18 Arabidopsis OFP family members, OFP2, OFP3, and OFP4 are the most closely related to OFP1 based on phylogenetic analysis of the Arabidopsis OFP gene family (Figure S1 in Supporting Information), but insertion mutants were only available for OFP1 and OFP4. Furthermore, while OFP4 is slightly up-regulated over the course of inflorescence stem development (Ehlting et al., 2005) OFP2 and OFP3 show no change. Thus, we chose to focus on OFP1 and OFP4 as potential in vivo interaction partners with KNAT7 that could control aspects of secondary wall formation.

We first re-tested the ability of KNAT7 to interact with OFP1 and OFP4 using the yeast two-hybrid system, in a targeted fashion rather than as part of a large matrix of interactions as previously reported (Hackbusch et al., 2005). Figure 1(a) shows that yeast cells expressing a KNAT7–DNA-binding domain (DBD) fusion and an OFP4–activation domain (AD) fusion interacted moderately well in this system, as judged by growth on both His (which detects weak interactions) and Ura (which detects stronger interactions) selective media, comparable to that of a positive control (MYB75–TT8 interaction; Zimmermann et al., 2004). Using similar criteria, a detectable but weak interaction was found between KNAT7 and OFP1. These data confirm and refine those of Hackbusch et al. (2005), suggesting that there are differences in the strength of interactions between KNAT7 and different OFPs.

Figure 1.

 Yeast two-hybrid assay of KNAT7–OFP (Ovate Family Protein) interactions.
(a) Top: diagram of constructs used in yeast two-hybrid assays. Bottom: assay of DNA-binding domain (DBD)–KNAT7 interaction with activation domain (AD)–OFP1 and AD–OFP4 using two reporter genes, URA3 (assayed on Ura medium) and HIS3 (assayed on His medium), with growth controls in Leu medium alone. Growth on His medium and absence of growth on Ura medium is indicative of weak protein–protein interactions. MYB75–TT8 interaction was used as a positive control, and DBD–KNAT7–AD empty vector interaction was used as a negative control.
(b) Top: representation of KNAT7 structure with MEINOX, KNOX, ELK, and homeodomains. Locations and names of KNAT7 fragments cloned separately and used for yeast two-hybrid and protoplast transfection assays are shown below. Middle: diagram of constructs used in yeast two-hybrid assays. Bottom: assay of KNAT7–OFP interacting domains and assay of DBD–KNAT7 fragment interaction with AD–OFP1 and AD–OFP4 using two reporter genes, URA3 (assayed on Ura medium) and HIS3 (assayed on His medium), with growth controls in Leu medium alone. Growth on His medium and absence of growth on Ura medium is indicative of weak protein–protein interactions. The MYB75–TT8 interaction was used as a positive control. The positive and negative controls are shown in part (a).
A schematic diagram of the yeast two-hybrid constructs tested in (b) is shown in Figure S2.

A previous report (Hackbusch et al., 2005) suggested that the homeodomain of certain KNOX proteins interacts with the OFP ovate domain to mediate protein–protein interactions between these classes of transcriptional regulators. To test this directly for the KNAT7–OFP1 and –OFP4 interactions, we generated a set of four fusions of KNAT7 domains, KNOX1, KNOX2, MEINOX, and Homeodomain, to the GAL4 DBD, and tested their abilities to interact with AD–OFP1 and –OFP4 fusions (Figures 1b and S2). The KNAT7 homeodomain interacted most strongly with OFP4 and OFP1, suggesting that the homeodomain is sufficient for interaction of KNAT7 with these OFP proteins.

To test for OFP–KNAT7 protein–protein interactions in planta, we employed the bimolecular fluorescence complementation (BiFC) technique (Hu et al., 2002; Tzfira et al., 2005; Shyu et al., 2006). OFP1 and OFP4 were fused to a C-terminal (C-EYFP) and KNAT7 to an N-terminal (N-EYFP) fragment of the enhanced yellow fluorescent protein (EYFP), neither being capable of fluorescence alone. Using an Arabidopsis leaf mesophyll protoplast transient expression system (Tiwari et al., 2006), we transformed different combinations of fusion constructs into protoplasts. Fusions of OFP1, OFP4 and KNAT7 to a complete EYFP generated fluorescence localized to the nuclei of transformed protoplasts (Figure 2a–c), consistent with the results reported by Wang et al. (2007) on nuclear localization of OFP1 and the results of Zhong et al. (2008) on KNAT7 nuclear localization. Co-transformation of truncated EYFP fusions to these genes revealed that co-expression of OFP1–C-EYFP with KNAT7–N-EYFP, as well as OFP4–C-EYFP with KNAT7–N-EYFP generated nuclear-localized fluorescence (Figure 2d,e). However, no fluorescence was detected when either the KNAT7–N-EYFP or OFP–C-EYFP construct was co-expressed with RACK1–C-EYFP and RACK1–N-EYFP, a non-interacting protein (RACK1, receptor for activated C kinase 1; Chen et al., 2006; Guo and Chen, 2008) that served as a negative control (Figure 2f; data not shown). These data indicate that both OFP1 and OFP4 can interact with KNAT7 in vivo.

Figure 2.

 Bimolecular fluorescence complementation analysis of OFP1, OFP4 and KNAT7 interactions in vivo.
Proteins fused to the full enhanced yellow fluorescent protein (F-EYFP) (a, b, c) and N- or C-terminal fragments (d, e, f) (N-EYFP or C-EYFP, respectively) were co-expressed transiently in protoplasts. Yellow fluorescence indicating reconstitution of YFP was analyzed by confocal laser scanning microscopy.
(a) A protoplast transfected with 35S:KNAT7–F-EYFP. Left: differential interference contrast (DIC) image. Middle: image with YFP channel. Right, merged images.
(b) Images of a protoplast transfected with 35S:OFP1–F-EYFP.
(c) Images of a protoplast transfected with 35S:OFP4–F-EYFP.
(d) Images of co-expressing OFP1–C-EYFP and KNAT7–N-EYFP; yellow fluorescence indicates that the proteins interact in vivo in the nucleus.
(e) Images of co-expressing OFP4–C-EYFP and KNAT7–N-EYFP indicate that the proteins interact in vivo in the nucleus as well.
(f) Images of co-expressing the construct KNAT7–N-EYFP with negative control RACK1–C-EYFP revealed no fluorescence.

KNAT7 and OFP4 are transcriptional repressors

Previous data indicated that OFP1 functions as a transcriptional repressor (Wang et al., 2007). To test whether OFP4 could also function similarly, we employed a protoplast transfection system (Wang et al., 2007), illustrated in Figure 3a. This assay system, employing a potent LexA DNA-binding domain–VP16 transcriptional activator and various Gal4 DNA-binding domain (GD)–effector fusions, has been well documented in its ability to assay transcriptional repressors as GD fusions, for example AUX/IAA proteins (Tiwari et al., 2004) and OFP1 (Wang et al., 2007). The β-glucuronidase (GUS) gene under the control of both LexA and Gal4 DNA-binding domain (DBD) sites (LexA[2×]–Gal4[2×]:GUS reporter gene) was used as a reporter, and was co-transfected with two effector plasmids. The first contains a chimeric protein consisting of the LexA DBD fused to the herpes simplex virus VP16 transcription activation domain (LD–VP16), driven by the CaMV 35S promoter. A second contains a chimeric gene encoding a protein consisting of the Gal4 DBD fused to the gene of interest, such as OFP4 (GD–OFP4), also under control of the CaMV 35S promoter. A control second effector plasmid contains a 35S-driven GD domain only.

Figure 3.

 OFP4 and KNAT7 are transcriptional repressors that interact in vivo.
A protoplast transfection system was used to assay the ability of OFP1, OFP4 and KNAT7 to repress VP16-activated transcription of the GUS reporter gene.
(a) Top: effector and reporter constructs used in the transfection assays. Bottom: GUS activity in protoplasts derived from Arabidopsis rosette leaves co-transfected with an LD-VP16 transactivator plasmid and one of three effector plasmids (GD-OFP1, GD-OFP4, or GD alone) together with the reporter plasmid. The expression of 35S:luciferase (Luc) was used to normalize the expression of the GUS reporter gene. Error bars represent the standard deviation of three replicate transfections.
(b) Top: effector and reporter constructs used in the transfection assays. Bottom: GUS activity in protoplasts derived from Arabidopsis rosette leaves co-transfected with an LD-VP16 transactivator plasmid and different combinations of effector plasmids (HA-OFP1, HA-OFP4, GD-KNAT7, or GD alone) together with the reporter plasmid. The expression of 35S:luciferase (Luc) was used to normalize the expression of the GUS reporter gene. Error bars represent the standard deviation of three replicate transfections.

Co-transfection of the LD–VP16 transactivator gene and an effector gene encoding only the Gal4 DBD (GD) resulted in activation of the GUS reporter gene (Figure 3a), but co-transfection with either the GD–OFP1 or GD–OFP4 effectors resulted in a strong repression of GUS activity (Figure 3a), demonstrating that, like OFP1 (Wang et al., 2007), OFP4 functions as a transcriptional repressor. Figure 3(b) shows that co-transfection of LD–VP16 with a plasmid containing a GD–KNAT7 fusion also resulted in a strong repression of the expression of GUS activity, relative to protoplasts transfected with the GD control, indicating that KNAT7 is also a transcriptional repressor. To determine if in vivo interactions between KNAT7 and OFP1 and/or OFP4 affect KNAT7 transcriptional activity, we fused both OFP1 and OFP4 to an N-terminal hemagglutinin (HA) tag and placed the genes under the control of the CaMV 35S promoter (HA–OFPs; Figure 3b). Plasmids containing these fusions were then co-transfected with the GUS reporter plasmid (LexA[2×]–Gal4[2×]:GUS), an effector plasmid containing the activator fusion gene (LD–VP16), and a second effector plasmid containing the GD–KNAT7 fusion protein. Figure 3(b) shows that both OFP1 and OFP4 enhanced the repression of LexA(2×)–Gal4(2×):GUS reporter gene expression relative to KNAT7 repression activity alone. This supports the binding of OFP1 and OFP4 to KNAT7 in vivo in this protoplast transient assay system and that this interaction enhances KNAT7 transcriptional repression.

To investigate which KNAT7 domains are required for the transcriptional repression function, KNAT7 effector gene fusions consisting of the GD fused to truncated versions of KNAT7 shown in Figure 1b were generated and tested for ability to repress of VP16-activated GUS reporter gene expression in the protoplast co-transfection assays (Figure 4a). A fusion of a KNAT7 N-terminal fragment containing only KNOX1 to GD (GD–KNAT7-KNOX1) failed to repress expression of the reporter gene. However, effector gene fusions containing either the KNOX2 domain (GD–KNAT7-KNOX2) or MEINOX domain (GD–KNAT7-MEINOX) caused significant repression of the VP16-activated transcription (Figure 4b), suggesting that the KNOX2 domain within the MEINOX domain is involved in repression. Finally, a fusion of the KNAT7 homeodomain region including the ELK motif to GD (GD–KNAT7-homoedomain) also caused repression, at a level near to that of the full fragment KNAT7 protein (Figure 4b). These results indicate that there are two major transcriptional repression domains within KNAT7, one within the fragment containing the KNOX2 domain and adjacent C-terminal sequences, and one within the C-terminal portion of the protein including the ELK motif and homeodomain.

Figure 4.

 KNAT7 transcriptional repression domains.
A protoplast transfection system used to assay for the location of the KNAT7 transcriptional repression domains.
(a) Schematic diagram of KNAT7 with four different truncated versions, KNOX1, KNOX2, MEINOX (KNOX1 + KNOX2), and Homeodomain.
(b) GUS activity in protoplasts derived from Arabidopsis rosette leaves co-transfected with an LD-VP16 transactivator plasmid and one of the effector plasmids together with the reporter plasmid. The expression of 35S:luciferase (Luc) was used to normalize the expression of the GUS reporter gene. Error bars represent the standard deviation of at least three replicate transfections.

Co-expression of OFP4/1 and KNAT7

KNAT7 is expressed in developing fibers and vascular bundles in inflorescence stems, in accordance with its role in secondary wall formation (Zhong et al., 2008). Furthermore, promKNAT7:GUS fusion expression is closely associated with the vascular system and interfascicular fibers in Arabidopsis seedlings, roots, and inflorescence stems (Li, 2009). To test the tissue and organ expression pattern specified by the OFP4 promoter, a 643-bp genomic fragment upstream of the OFP4-coding region was fused to the GUS reporter gene to generate promOFP4:GUS for plant transformation. Four independent transgenic Arabidopsis lines were analyzed for GUS activity using the histochemical assay. Consistent expression patterns were found in all four lines, and representative expression is shown in Figure 5.

Figure 5.

 Histochemical assay of OFP4:GUS, OFP1:GUS, and KNAT7:GUS expression.
(a–e) Expression of promOFP4:GUS in 7-day-old seedlings: (a) cotyledon, (b) junction of hypocotyl and root, (c) root, (d) whole seedling, (e), root tip.
(f, g) Expression of promOFP4:GUS in inflorescence stem cross-sections.
(h, i) Expression of promOFP1:GUS in inflorescence stem cross-sections at higher magnification.
(j, k) Representative expression of promKNAT7:GUS in cotyledons (j) and an inflorescence stem cross-section (k).
Bars: 20 μm for a, b, c, e, h, j, k and 40 μm for g, i.

In 7-day-old seedlings (Figure 5d), promOFP4:GUS was expressed mainly in the root, at the root–hypocotyl junction, and in cotyledons (Figure 5a–c). In roots, expression was especially strong in the vascular cylinder (stele) (Figure 5c) and at the root tip (Figure 5e), while in cotyledons, expression was observed in both veins and other tissues (Figure 5a). In mature plants, promOFP4:GUS activity in cross-sections of inflorescence stems was strong in xylem as well as in the cortex adjacent to interfascicular fiber cells (Figure 5f,g).

In accordance with the previously reported tissue and organ expression patterns of the 1383-bp promOFP1:GUS reporter construct (Wang et al., 2007), we found highest levels of promOFP1:GUS expression in the roots of seedlings, especially in the vascular cylinder, of the representative lines assayed (data not shown). Expression of promOFP1:GUS expression in cross-sections of inflorescence stems has not been reported, and here GUS activity was detected in xylem and the cortex cells close to interfascicular fibers (Figure 5h,i), similar to that of promOFP4:GUS.

For comparison, the GUS expression pattern of a promKNAT7:GUS fusion described in detail (Li, 2009) is shown in Figure 5j,k. While the 2-kb KNAT7 promoter employed in promKNAT7:GUS directs expression more specifically to the vascular system than the OFP4 promoter (compare Figure 5d,e with Figure 5j), the patterns of OFP1 and OFP4 promoter activities overlap with that of KNAT7. In addition, the patterns specified by all three promoters closely coincide in the inflorescence stem, suggesting that these genes are expressed in common in certain cell types.

OFP4 loss-of-function mutant phenocopies that of knat7

To further investigate the potential roles of OFP1 and OFP4 in regulating secondary cell wall formation as interacting partners with KNAT7, we analyzed the phenotypes of OFP1 and OFP4 loss-of-function mutants. The OFP1 mutant ofp1-1 has been characterized previously, and is morphologically indistinguishable from a wild-type plant (Wang et al., 2007). We obtained two ofp4 T-DNA insertion alleles. ofp4-1 (SALK_014905) was previously described but not characterized (Pagnussat et al., 2007). We confirmed that ofp4-1 contains a T-DNA insertion located in the promoter region 109 bp upstream of the start codon (Figure 6a) and that OFP4 RNA is not detectable in ofp4-1 (Figure 6b). We obtained another T-DNA insertion line, SALK_022396, with a T-DNA insertion located 274 bp downstream of the OFP4 start codon, designated ofp4-2 (Figure 6a). A RT-PCR analysis confirmed that a full OFP4 transcript was undetectable in the homozygous line, indicating that ofp4-2 is also a loss-of- function mutant allele of OFP4 (Figure 6b). We used ofp4-2 for most of the studies. At the rosette stage, the morphology of ofp4-2 homozygous plants was normal compared with wild type Col-0, similar to homozygous ofp1-1 plants (Figure S2). We carried out phenotypic analyses of the anatomy of cross-sections taken from the bases of ofp1-1, ofp4-1 and ofp4-2 inflorescence stems (Figures 7 and S3). No differences in xylem or interfascicular fiber morphology were evident in homozygous ofp1-1 relative to the wild type (Col-0) (Figure 7b), but sections taken from ofp4-1 and ofp4-2 stems revealed an irx phenotype (Figures S3B and 7c).

Figure 6.

 Analysis of OFP1 and OFP4 insertion mutants.
(a) Transposon and T-DNA insertion sites in OFP4 (ofp4-1 and ofp4-2). The shaded box represents translated region and the open box represents the 5′ untranslated region.
(b) The RT-PCR analysis of OFP4 expression in wild-type Col-0; ofp4-2 mutant and 35S:OFP4 plants. Actin was used as a control.

Figure 7.

 Phenotypic characterizations of loss-of-function mutants of OFP1, OFP4 and the ofp1/ofp4 double mutant.
Cross-sections of the basal internodes of 6-week-old wild type (Col-0) (a), ofp1 (b), ofp4 (c) ofp1/ofp4 double mutant plants (d). The stars indicate the irx morphology in vessels; arrows show the general locations of interfascicular fiber (IFF) cell wall thickness measurements indicated below (= 40). Bars: 40 μm.

Closer examination by TEM showed more clearly that vessel elements in ofp4-2 vascular bundles were collapsed and irregularly shaped compared with those of the wild type (Figure 8a,b,e,f). Although normal in morphology, the walls of ofp4-2 interfascicular fiber cells appeared to be thicker than those in wild-type fiber cells (Figure 8c,d,g,h). Quantitative measurements of wall thickness taken from TEM micrographs (= 60 cells) showed that ofp4-2 interfascicular fiber cell walls were slightly thicker than those of the wild type, while vessel and xylary fiber cell walls were thinner (Table 1). Analysis by Student’s t-test indicated that the quantitative differences between wild-type and mutant cell wall thicknesses in all data sets were statistically significant (< 0.01). Therefore, the ofp4 loss-of-function phenotype in the inflorescence stem (irx, increased interfascicular fiber wall thickness) resembles that of knat7 mutants (Brown et al., 2005; Li, 2009). These data suggest a role for OFP4 in secondary wall formation function.

Figure 8.

 Transmission electron micrograph of the secondary walls in the wild type (andash;d) and ofp4-2 mutant (endash;h).
(a–d) sections from a wild-type plant.
(e–h) Sections from a ofp4-2 plant.
(a, b, e, f) Showing the collapsed vessels of ofp4 and (c, d, g, h) show the thicker interfascicular fiber walls in ofp4 plants: iff, interfascicular fiber; ve, vessel; xy, xylary fiber. Bars: 2 μm.

Table 1.   Secondary cell wall thickness in wild-type and ofp4-2 stems
GenotypeInterfascicular fiber wall thickness (μm)aVessel wall thickness (μm)aXylary fiber wall thickness (μm)a
  1. aData are means ± SE from at least 60 cells measured from transmission electron microscopy micrographs of the bases of primary inflorescence stems.

Wild type1.48 ± 0.370.83 ± 0.260.97 ± 0.32
ofp4-21.77 ± 0.370.61 ± 0.220.75 ± 0.21

OFP1 and OFP4 are very closely related among the 18 Arabidopsis OFP proteins (Figure S1), and functional redundancy has been suggested among OFP family members (Wang et al., 2007). Therefore, an ofp1 ofp4 double mutant was generated by crossing ofp1-1 with ofp4-2. The double mutant had wild-type morphology at the rosette stage (Figure S2). Sections from the bases of inflorescence stems showed a collapsed vessel (irx) phenotype, similar to the single ofp4-2 mutant phenotype (Figure 7d), but no apparent additive phenotype was observed relative to the ofp4-2 single mutant.

To confirm that the T-DNA insertion in OFP4 is indeed responsible for the phenotypes observed in the ofp4-2 mutant, we generated a promOFP4:OFP4–GFP with a native promoter construct transformed into ofp4-2 mutant plants for a complementation analysis. The representative line showed ofp4 mutant phenotypes recovered in cross-section in inflorescence stems of overexpression plants (Figure S4).

OFP4/1 and KNAT7 in plant development

To further test the roles of OFP1, OFP4 and KNAT7 in a KNOX–OVATE complex regulating secondary wall formation, we generated double mutants by crossing a knat7-1 mutant line with ofp1-1 and ofp4-2. Double mutants identified in the F2 population by PCR-aided genotyping had no morphological differences compared with the wild type. Light microscopic analysis of cross-sections from inflorescence stems of double mutants revealed that both of knat7 ofp1 and knat7 ofp4 double mutants exhibited similar phenotypes to knat7, with collapsed vessels and increased thickness of interfascicular fiber cell walls (Figure 9b,c). The lack of an additive phenotype is consistent with the view that KNAT7 and OFP1/4 may function in the same pathway.

Figure 9.

 Anatomical characterization of knat7/ofp1 and knat7/ofp4 double mutants.
Images show cross-sections of internodes at the bases of 8-week-old plants: (a) wild type Col-0, (b) knat7/ofp1 double mutant, (c) knat7/ofp4 double mutant. Arrows show vessels with irx morphology. Bars: 40 μm.

To test whether OFP1 and OFP4 functions are manifested through a KNOX–OVATE regulatory complex containing KNAT7 and thus require functional KNAT7, we generated double mutants by crossing a knat7-1 mutant line to 35S:OFP1 (Wang et al., 2007) and 35S:OFP4 (Figure 6b) overexpression lines. The 35S:OFP1 overexpression phenotype has been described previously (Wang et al., 2007). The pleiotropic phenotype of a representative 35S:OFP4 line is shown in Figures 10 and S5, and was similar to the 35S:OFP1 overexpression phenotype but more severe, including dwarfism, ovate-shaped organs, and reduced fertility.

Figure 10.

 Phenotypes of transgenic plants overexpressing 35S:OFP4 in Arabidopsis.
(a) Ten-day-old wild-type (Col-0) and 35S:OFP4 seedlings. (b) Rosette of wild-type (left) and 35S:OFP4 (right) plants. (c) Eight-week-old 35S:OFP4 plants. (d) The morphology of siliques of 35S:OFP4 plant (left) compared with a wild-type silique (right). (e) Flower buds in the wild type (left) and 35S:OFP4 (right). (f) Cross-section from the basal internode of an 8-week-old wild-type (Col-0) plant. (g) Cross-section from the basal internode of an 11-week-old 35S:OFP4 plant. Bars: 100 μm for f, and g.

At the seedling stage, formation of ovate-shaped cotyledons and leaves by OFP4 overexpression was suppressed in the knat7 background (Figure 11a). At the rosette stage, suppression of the OFP1 and OFP4 overexpression phenotype in knat7, relative to the KNAT7 background, was more pronounced, with both double mutants exhibiting wild-type morphology (Figure 11b). Further analysis revealed that mature 35S:OFP1 knat7 and 35S:OFP4 knat7 double mutant plants were similar in morphology to wild-type plants, with normal inflorescence stem development (Figure S6), and retained the knat7 irx phenotype as expected (Figure S7).

Figure 11.

 Phenotypic characterization of 35S:OFP1 knat7, and 35S:OFP4 knat7 plants in seedlings and rosette stages.
(a) Morphology of seedlings transgenic for 35S:OFP4 in the knat7 background relative to the wild-type (KNAT7 KNAT7) background.
(b) Morphology of rosette stage for 35S:OFP1 and 35S:OFP4 in the knat7 background compared with the wild type.


Previous data indicated that some members of the Arabidopsis OFP family interact with certain KNOX and BLH TALE homeodomain proteins via protein–protein interactions (Hackbusch et al., 2005). Since OFP proteins themselves do not contain predicted DNA-binding domains, and OFP1 acts as a transcriptional repressor (Wang et al., 2007), a model for the function of OFPs in developmental control is that their interactions with KNOX and/or BLH proteins result in KNOX–BHL–OFP complexes that repress transcription (Pagnussat et al., 2007; Wang et al., 2007). In this study, several lines of evidence support the interaction of OFP1 and OFP4 with KNAT7 to form a transcription repression complex, and suggest that a OFP4–KNAT7 complex regulates secondary wall formation in Arabidopsis.

KNAT7 interacts with OFP1 and OFP4 in vivo to form transcription repression complexes

We used yeast two-hybrid assays to verify that KNAT7 interacts with OFP1 and OFP4, but the KNAT7–OFP1 interaction appeared much weaker. However, based on BiFC results after protoplast transfection, KNAT7–OFP1 and KNAT7–OFP4 interactions appeared equally strong (Figure 2), and both OFP1 and OFP4 were able to enhance KNAT7 repression activity in the protoplast system (Figure 3b), consistent with biologically meaningful in vivo interactions. It is possible that additional factors, such as BLH TALE homeodomain proteins known to interact with KNOX proteins (Bellaoui et al., 2001) and KNAT7 (Hackbusch et al., 2005) are present in protoplasts, stabilizing OFP–KNAT7 interactions in vivo, and thus resulting in stronger OFP1–KNAT7 interaction than observed in protoplasts.

When recruited to a promoter activated by the LexA:VP16 transcriptional activator, both the KNAT7 MEINOX domain, which is known to be required for heterodimerization with BLH proteins via their SKY and BELL domains (Bellaoui et al., 2001; Muller et al., 2001), and the KNAT7 homeodomain that functions by directly interacting with target DNA or other proteins (Gehring et al., 1994) were effective in transcriptional repression (Figure 4b). Both domains individually repressed transcription at levels similar to the full KNAT7 protein, and the MEINOX KNOX2 domain alone was nearly as effective. Thus, repression domains may be present in both the KNOX2 and homeodomain portions of the protein, or reside in the center of KNAT7 between KNOX2 and the homeodomain. Our yeast two-hybrid data showed that the OFP4 homeodomain also mediates interaction with KNAT7, consistent with results reported by Hackbusch et al. (2005) and with the hypothesis that the homeodomain of Arabidopsis BLH and KNOX proteins mediates protein–protein interactions with the conserved OVATE domain of OFP proteins. Since the apparent transcriptional repression domains identified in KNAT7 correspond to domains mediating protein–protein interaction with BLH and OFP proteins, we cannot exclude the possibility that the repression activity observed in protoplasts is indirect, due to in vivo recruitment of interacting proteins that act as repressors. Transfection assays using protoplasts from lines with mutations in genes encoding such interacting proteins could be used to address this point.

A KNOX–OVATE complex is involved in Arabidopsis secondary wall formation

Little functional information is available about TALE homeodomain protein–OFP complexes, with exception of a report that a BLH1–KNAT3–OFP5 complex regulates egg development in Arabidopsis (Pagnussat et al., 2007). Given that KNAT7 plays a role in secondary cell wall and vascular development (Brown et al., 2005; Zhong et al., 2008; Li, 2009), the in vivo interaction between KNAT7 and OFP1 and OFP4 suggests that one or both of these OFP proteins could work with KNAT7 as a part of a KNOX–OVATE complex to regulate some aspects of secondary cell wall formation. Our finding that KNAT7 and KNAT7–OFP1 and –OFP4 complexes are transcriptional repressors, rather than activators, in vitro suggests that their role may be to repress aspects of secondary cell wall formation, rather than to promote it. Thus, the increase in interfascicular fiber wall thickness in ofp4 (Figure 8) and knat7 insertion mutants (Li, 2009) may reflect increased commitment to secondary cell wall biosynthesis in these cells. The decrease in observed interfascicular fiber wall thickness phenotype in plants expressing a dominant negative KNAT7 variant (Zhong et al., 2008) may thus simply reflect an increase in the ability of KNAT7 to repress secondary cell wall formation in these cells.

Plants overexpressing OFP4 had similar pleiotropic phenotypes to those overexpressing OFP1 (Figures 10 and S5), which is in agreement with the view that OFP proteins including OFP4 may have overlapping functions in the regulation of multiple aspects of plant growth, and organ and cell morphogenesis. In contrast, the ofp4-2 loss-of-function mutant phenotype is much more subtle, and appears to be restricted to xylem and interfascicular fiber differentiation, suggesting that OFP4 may play a more specific role in the differentiation of these cells.

If they function together with KNAT7 as part of a regulatory complex in vivo, OFP1 and OFP4 would be expected to have expression patterns that overlap with KNAT7. The OFP1 promoter had previously been shown to be expressed at the root and root–hypocotyl junction and in multiple organs (Wang et al., 2007), but its cell-type-specific expression was not known. Our analysis (Figure 5) showed that promOFP1:GUS and promOFP4:GUS expression is strong in the root vascular cylinder (stele) and in the inflorescence stem, where expression in developing xylem and the cortex immediately adjacent to interfascicular fibers was predominant. This is in accordance with the expression pattern of the promKNAT7:GUS expression pattern (Figure 5; Li, 2009) indicating that OFP1 and OFP4 proteins are probably present in the same cell types as KNAT7.

A biological function of OFP4 in regulating secondary cell wall formation through its interaction with KNAT7 is further supported by the phenotypes of the ofp4-2 single mutant and the ofp1-1ofp4-2 double mutant. Both mutants exhibited inflorescence stem phenotypes similar to that of knat7 (Brown et al., 2005; Li, 2009) with characteristic irx phenotypes and enhanced thickness of interfascicular fiber cell walls (Figures 7 and 8). However, we failed to observe any irx or other secondary cell wall phenotype in the ofp1-1 mutant (Figure 7). This suggests that although OFP1 can interact with KNAT7 in vivo and shares a similar expression pattern, it may not function together with KNAT7 in a KNOX–OVATE complex regulating secondary cell wall formation. The weak interaction between OFP1 and KNAT7 in yeast two-hybrid assays (Figure 1) further suggests that KNAT7–OFP1 interaction could be distinct from OFP4–KNAT7 interaction. Alternatively, since the OFP1 homologs OFP2 and OFP3 (Figure 1) have also been shown to interact with KNAT7 in yeast two-hybrid assays (Hackbusch et al., 2005), functional redundancy between OFP1 and OFP2 and/or OFP3 could mask the ofp1-1 mutant phenotype. These data suggest, however, that OFP4 may play a more important or specialized role in developmental regulation through its interaction with KNAT7.

Both knat7ofp1 and knat7 ofp4 double mutants displayed similar phenotypes in inflorescence stems to the knat7 mutant, i.e. irx and thicker interfascicular fiber cell walls (Figure 9). The lack of an additive phenotype indicates that KNAT7 and OFP4 might work in the same regulatory pathway, consistent with their postulated roles in a KNOX–OVATE regulatory complex.

Evidence for the existence of biologically active KNOX–OVATE complexes involving both OFP1 and OFP4 and KNAT7 also comes from the fact that KNAT7 appears to be required for the pronounced and pleiotropic phenotypic effects of OFP1 and OFP4 overexpression. In the knat7 mutant background, these phenotypes were rescued (Figures 11 and S6) and the inflorescence stems phenotypes of these double mutants exhibit knat7-like phenotypes (Figure S7). Thus, both of these OFPs might work primarily in a complex requiring KNAT7, and the regulatory activities might extend beyond secondary cell wall formation. Overexpression of OFP1, for example, has been shown to have general effects on cell growth and to affect GA homeostasis (Wang et al., 2007). Since KNAT7 expression seems to be primarily restricted to cells with developing secondary walls (Figure 5; Zhong et al., 2008; Li, 2009), in contrast to the broader expression patterns of OFP1 and OFP4 (Wang et al., 2007; Figure 5), the mechanism by which KNAT7–OFP1/4 interactions pleiotropically impact broader plant developmental processes is unclear and will be the subject of future investigations.

In summary, our data support a model in which KNAT7–OFP1 and –OFP4 protein–protein interactions enhance KNAT7-mediated transcriptional repression of target genes, and suggest that KNAT7–OFP4 interaction is of particular importance in the regulation of secondary cell wall formation. The KNAT7–OFP regulatory module might also include unknown BLH proteins, and work in a similar manner to other functional modules linking TALE homeodomain proteins and OFPs, such as the postulated interaction of OFP1 with KNAT1 to repress GA20ox1 expression (Wang et al., 2007), and the role of OFP5 in negatively regulation of a BLH-KNOX during early embryo sac development (Pagnussat et al., 2007).

Experimental Procedures

Plant materials

Arabidopsis thaliana Heynh ecotype Columbia (Col-0) was used as the wild type, and all mutants and transgenic lines were in this background. Seedlings were grown from seeds sterilized and sown on half-strength Murashige and Skoog (MS) basal medium with vitamins (PlantMedia, http://www.plantmedia.com/), then cold-treated at 4°C in the dark for 48 h, and grown at 22°C with a 16/8 h (light/dark) photoperiod at approximately 120 μmol m−2 sec−1 light for 7–10 days.

The T-DNA insertion mutant allele of OFP4, SALK_022396, designated as Atofp4-2 was identified using SIGnal database (http://signal.salk.edu/) and seeds were obtained from the Arabidopsis Biological Resources Center (ABRC, http://abrc.osu.edu/). An OFP4 gene-specific primer (Table S1) and the T-DNA-specific primer JMLB1 (5′-GGCAATCAGCTGTTGCCCGTCTCACTGGTG-3′) were used for PCR genotyping. The T-DNA insertion site was verified by sequencing. Double mutants were generated by crossing the two individual homozygous lines, and double mutants were identified in the F2 generation by PCR-aided genotyping.

Wild-type Col-0 plants were used for transformation by Agrobacterium tumefaciens GV3101 using the floral dip method (Clough and Bent, 1998). Phenotypes of transgenic plants were examined in the T1 generation, and confirmed in T2 to T4 generations. For all transgenic plants, at least five transgenic lines with similar phenotypes were obtained. Materials for study of Arabidopsis inflorescence stem development were grown in soil until the inflorescence stems had two or three fully expanded siliques (6–8 weeks). Stem fragments 1 cm from the bottom were sampled for phenotypic analysis.

Plasmid constructions and Arabidopsis protoplast transfection assays

The generation of 35S:HA-OFP1, 35S:GD-OFP1 and OFP1prom:GUS constructs has been described previously (Wang et al., 2007). To generate the 35S:HA-OFP4 and 35S:GD-OFP4 constructs, the full-length open-reading frame (ORF) of the OFP4 gene was amplified by PCR using genomic DNA isolated from 10-day-old, light-grown Arabidopsis seedlings, because the OFP4 gene contains a single exon. The PCR fragment was then cloned in frame with a N-terminal HA epitope tag or a GD tag into the pUC19 vector under the control of the double 35S enhancer promoter of CaMV (Tiwari et al., 2003; Wang et al., 2005). Corresponding constructs with the HA tag in the pUC19 vector were digested with a restriction enzyme, EcoRI, then sub-cloned into binary vector pZP211 for plant transformation (Hajdukiewicz et al., 1994) 35S:GD-OFPconstructs in the pUC19 vector were used for plasmid DNA isolation and protoplast transfection. The OFP4 promoter (a 643-bp fragment immediately before the start codon of the OFP4 gene) was amplified by PCR and replaced the OFP1 promoter in OFP1prom:GUS to generate the OFP4prom:GUS construct. Effector OFP1 and OFP4 genes, transactivator LD-VP16 and reporter LexA()-Gal4():GUS were as described by (Wang et al., 2007). All reporter and effector plasmids used in transfection assays were prepared using the EndoFree Plasmid Maxi kit (Qiagen, http://www.qiagen.com/). Protoplast isolation and transfection were performed as described previously (Tiwari et al., 2006; Wang et al., 2005). Statistical significance tests were performed with Student’s t-test using tools at http://www.graphpad.com/quickcalcs/ttest1.cfm. All transfection assays were performed as at least three replicates, and assays were repeated on at least two separate occasions.

GUS expression assay

The GUS activity was assayed in 7-day-old seedlings and the fresh-cut hand-sections from inflorescence stems of 6–8-week-old plants by incubating tissues in a solution containing 100 mm sodium phosphate buffer, pH 7.0, 0.1% Triton X-100, 1 mm substrate 5-bromo-4-chloro-3-indolyl-β-d-glucuronide (X-Gluc; Rose Scientific Ltd, http://www.rosesci.com), and 0.5 mm potassium ferricyanide at 37°C for 1–12 h. Tissues were cleared with 70% ethanol and placed in 50% glycerol for analysis. All light microscopic observations were performed with an Olympus AX70 bright field microscope (http://www.olympus.com/).

RNA extraction and reverse transcription (RT)-PCR

RNA samples treated with RNase-Free DNase (Qiagen) were extracted from fresh tissue using the Qiagen RNeasy Plant Mini kit according to the manufacturer’s instructions. Two micrograms of total RNA was used for reverse transcriptase synthesis using the Omniscript RT kit (Qiagen) according to the manufacturer’s instructions. The ACTIN (ACT/At2g37620) gene was used as a control for RT-PCR.

Yeast two-hybrid assays

Clones of KNAT7, OFP1 and OFP4 ORFs were isolated from Arabidopsis cDNA, truncated KNAT7 regions corresponding to the KNOX1, KNOX2, MEINOX and homeodomains were amplified from a KNAT7 plasmid clone and were introduced into a Gateway® entry vector pCR8 (Invitrogen, http://www.invitrogen.com/). Each clone was transferred to Invitrogen Gateway® compatible yeast two-hybrid bait and prey vectors by using LR cloning. The interactions between OFP1/OFP4 and KNAT7 were tested by using ProQuest Two-Hybrid System (Invitrogen) as described previously (Guo et al., 2009). KNAT7, KNOX1 domain, KNOX2 domain, MEINOX domain and homeodomain fragments were cloned into a bait vector (pDEST32) to generate N-terminal GAL4 DBD fusions, and OFP1 and OFP4 were cloned into the prey vector pDEST22 to generate N-terminal GAL4 DNA AD fusions. The known interaction between MYB75 and TT8 (Zimmermann et al., 2004) was used as a positive control. The interaction between KNAT7 and empty prey vector was used as a negative control. Positive interactions were identified based on their ability to activate the HIS3 and URA3 reporter genes, as described (Guo et al., 2009).

Bi-molecular fluorescence complementation (BiFC) analysis

The cloning vectors pSAT6-EYFP-N1, pSAT6-EYFP-N1 (N-terminal of EYFP) and pSAT4A-cEYFP-N1 (C-terminal of EYFP) (Tzfira et al., 2005) were used to express fusion proteins under the control of the CaMV 35S promoter. We introduced XhoI and HindIII restriction sites into PCR primers used to amplify KNAT7, OFP1 or OFP4 ORFs (primers in Table S1) and cloned restriction enzyme-digested fragments into the pSAT6-EYFP vectors. Fusions to N-terminal and C-terminal truncated EYFP variants were generated in the same manner. For transient expression using Arabidopsis leaf protoplasts, 10 μg of plasmid DNA was transfected into protoplasts, and incubated for 20–22 h. After incubation, cells were concentrated by centrifugation. Two or three drops of cell solution was deposited on a microscope slide and between two cover slips, and covered with another cover slip, to provide space to protect protoplasts from damage. Slides were examined immediately using a Leica DM-6000B fluorescent microscope (http://www.leica.com/). The YFP excitation was examined and photographed using phase and differential interference contrast (DIC) and a Leica FW4000 digital image acquisition and processing system (Leica Microsystems).


All micrographs were acquired digitally. Adobe Photoshop (http://www.adobe.com/) and ImageJ software (http://rsb.info.nih.gov/ij/index.html) were used for image processing. Stem sections about 200 μm thick were hand-cut and stained in aqueous 0.02% toluidine blue O (Sigma, http://www.sigmaaldrich.com/) for 1–2 min, rinsed briefly in distilled water, and mounted in water. Samples were viewed using an Olympus AX70 light microscope. For tissue embedding and for transmission electron microscopy (TEM), tissue was taken 1 cm from the base of inflorescence stems from 6-week-old plants. The TEM images were viewed on Hitachi H7600 PC-TEM (http://www.hitachi.com/) at an accelerating voltage of 80 kV. Photographs were taken using an ATM Advantage HR digital CCD camera (Advanced Microscopy Techniques, http://www.amtimaging.com/). Cell wall thickness measurements were taken from TEM micrographs using ImageJ from 60 cells per genotype, at standardized positions. Analysis by Student’s t-test was carried out using tools at http://www.graphpad.com/quickcalcs/ttest1.cfm.

Accession numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative database under the following accession numbers: At1g62990 (KNAT7; IRX11), At5g01840 (OFP1), and At1g06920 (OFP4).


We thank Michael Friedmann and Bjoern Hamberger (University of British Columbia, Canada) for helpful discussions and advice, the UBC Bioimaging Facility for technical assistance, and the ABRC and The Arabidopsis Information Resource (TAIR; Carnegie Institute of Washington, Stanford, CA, USA) for seeds and biological information, and Apurva Bhargava (University of British Columbia, Canada) for discussions and control yeast two-hybrid constructs. This work was supported by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grants to CJD and J-GC, and by the NSERC Green Crops Network (funds to CJD). EL was partially supported by a University of British Columbia University Graduate Fellowship.